5x10 servo router rebuild - analog of step/dir with mesa card?

warreng wrote:

Pardon my ignorance... Based on your comment I presume that the servo tuning is stored in the servo drives?

I have had the understanding that I DO want the LinuxCNC feedback loop in order to make a [more] closed loop system which is more better than having the control completely blind to the position (or loss of position) of the motors.

Seen this after i already replied above.
Yes, in position mode the loop is closed in the drives, so the tuning is done on the drives.
Since LinuxCNC can do steppers with feedback and loop closed in LinuxCNC, it is much better to use the drives as is as they are already tuned there, and tuning steppers with feedback is much, much easier than analog servos.
Another good thing is there is no fear of motors running at full speed if any mistake is made during wiring and tuning in analog mode as the drives have no control over motor speed in that mode, they do whatever is required from the analog input.

I'm back with a couple more questions about this. 

A) I went down the rabbit hole recently wrt s curve trajectory planning and have been looking at other control systems besides LinuxCNC just to be sure another path might not be better. I'm astounded that practically all the other options I've found are windows only. I'm not thrilled with that on principle alone. But I'm sure that they are all reasonable when set up correctly. My first question is whether s curve is even desirable for a 5000 lb 5'x10' CNC wood router with helical rack and servo motors that can do about 1500 ipm on X&Y OR if the look ahead and buffering in LinuxCNC with the current trap planner is likely to be just fine for this machine. It seems that a lot of LinuxCNC users are running on metal cutting mills/lathes/plasma and/or small desktop routers so I'm looking for a reality check for this application.

It's been my assumption based on this thread and others that the feedback loop in LinuxCNC can close in two ways. Clearly, with servos and encoders, it can close the "servo loop" and LinuxCNC is in full charge of the drives/motors. This means tuning in LinuxCNC, if I understand correctly. But can someone clarify how it works and what LinuxCNC actually does with encoder data when the tight servo loop is closed in the drives instead? Is there a "looser" loop of some sort? 

C) Any general or specific comments about 1) Centroid vs 2) WinCNC vs 3) Auggie with a PoKeys vs 4) CSMIO vs 5) LinuxCNC for this wood router? I have the opportunity to steal a WinCNC control from another machine; would anyone recommend that over LinuxCNC for this application? 

I've been using machine's at work with WinCNC and it's DOS based predecessor for more than 20 years, and Linuxcnc for a bit over half that time. Personally I would not choose it over Linuxcnc for a new machine build. It works ok, but its configurability is far more limited than Linuxcnc. I also don't fully trust the motion planner in WinCNC. I'm not sure that it really does proper acceleration or deceleration profiles, and it certainly doesn't do jerk limited motion. (If a machine set up with WinCNC is actually doing jerk limiting, it is being done in the servo drive, in spite of what ever WinCNC is commanding them, not because of it.)

Linuxcnc's dialect of G-code seems to be more "normal" and the canned cycles are more Fanuc like than WinCNC's g-code dialect. Also writing macros and subprograms seems easier in Linuxcnc. One think that is easier in Wincnc is setting up repeat arrays, but it uses a propriatary G-code that is unique to Wincnc.(L10).

Also if you ever have trouble with your Wincnc system, and need support, you will likely be stuck with having to pay $3000-5000 to upgrade to the latest supported version. (Has happened to us a couple of times.)

We've had 7 different routers that started out with their software. Only one is still running Wincnc. (The rest are now running Linuxcnc, were sold, or both.)

My googling shows hints that WinCNC has some s-curve support. But I'm not educated enough to know how complete or useful it is. Thank you for the first hand comments about WinCNC. 

Your comment about jerk limiting in the servo highlights a hope I had wrt the Yaskawa drives I have. But, alas, Sigma II doesn't appear to have this feature. I'm also confused about how a machine can keep synchronicity while drives are doing acceleration modifications in isolation. 

If I've read correctly, you are running primarily wood cutting machines? 

PCW wrote:

I would think that leaving it step/dir would be the easiest path
and would require few wiring changes. It also would not require
changing the drive setup, tuning the drives or tuning LinuxCNC's
feedback loop.

The only real advantage to an analog setup is that you can disable
the drives without losing position.
 

i have a working setup with mesa 7i95t (step/dir) and Delta B3-L drives. I am using velocity mode and feedback from drive/encoders, so position is kept while disabling the drives. That means you dont have to re-home after powering off, which is useful on long axes.
You have to tune servo drives in position mode first. Then you can switch to velocity mode and tune lixnucnc PID to minimize following error. I was able to reach the peak of ~ 0.005 during accelerations.

github.com/zmrdko/mesa_7i95t_config